AWS vs GCP vs Azure

I'm not a fan of any big cloud provider for indies. I don't want to spend any more time than necessary messing with servers and deployments. That's why I use PaaS services like Vercel, Railway or Supabase.

I often see these type of question on IH, so I thought I'd share my two cents worth. I always create Docker images for all my backend services so they are agnostic of any of the platforms.

If my frontend code is calling through an HTTP service using a REST endpoint then there is a clear decoupling of the frontend from the backend. I usually document this using an Open API defintion.

Most of the API begin their life in Google Cloud Run and I always make sure the service endpoints are hosted on my own domain. This way I can switch out the backend infrastructure without the frontend knowing or caring.

I’ve used AWS and GCP, I’ve not used Azure so I can’t comment.

AWS:
Great for enterprise who want to move from on prem to cloud and have a lot of flexibility.

Loads of services to choose from, they clone open source projects (sometimes badly) and put a badge on them

They often offer 2/3 versions of seemingly the same service which makes it difficult to figure out what to use sometimes.

I’ve found the setup to be complex when building secure scalable clouds. Even products like fargate, which are supposed to be managed, require a fair amount of setup for load balancers and VPCs.

Higher startup costs imo than GCP (but depends on the stack you use).

Lots of projects seem half finished, buggy UI

Easy to get lots of credits

Experience: I used to work at a company who used AWS and they spent 20k pm.

GCP:
More opinionated in how the cloud should work.

Lacking in features compared to AWS, but they have a great array of more modern cloud native feature. (Ie they don’t just clone open source projects and put a badge on them, they typically build from the ground up)

Cloud run is great and the best way to build something for low startup cost. A secure, scalable, HA stack with a Postgres db can be created for 30usd pm. (You could go at low as 20 I think if you didn’t have HA Postgres)

Harder to get credits and they tend to last less time (gcp gave us 2k which expire in 90 days which is totally pointless)

Discounts are automatically applied and can be shared across resources and accounts (I think). I believe AWS is more rigid in how they apply discounts, but I haven’t checked in a while.

Slow UI but not bugggy like AWS

Experience: building a startup using gcp for 7 months

——-

In short if you can get credits for gcp I would use them, I’ve found it far easier to setup and manage and I think cloud run is great and very cost effective.

I’ve never used supabase it looks ok, would be interested to see if it scales well in terms of pricing, features and performance.

Supabase

AWS is the standard. I may use Digital Ocean on indie projects due to pricing, but the stuff you learn from AWS generally carries into your career.

My product, a Terraform framework for Kubernetes supports AKS, EKS and GKE. So I am familiar with all three. I run my own stuff on GCP, mostly for subjective personal preference.

Thanks for sharing the information. please visit here to know more about aws, https://sites.google.com/view/awstrainingatsevenmentor/home

Well we use all and we provide a service that helps you use all.
At the moment we support AWS and have some features in preview for Azure but soon will finalize them.
Each of them provide some services that are better than the other ones.
Azure LogicOps or GCP Data studio are two examples.

I'm loyal to AWS, and won't change.

Google Cloud is really a distant third in this race. Azure so far just copies AWS but I'm not up to date enough to know if they broke out on any particular feature.

What is your stack?